Question Concerning Technical Details of Cisco Voice RTP and What It Requires To Deliver Good Quality Voice

1) Our Company's "voice guy" CCIE candidate has convinced IT management that Cisco voice either from a hard phone or from a jabber soft phone can only deliver business quality if the call setup and rtp call payload for station to station internal calls is all 100% routed across MPLS / BGP.
a. Assertion is that AT&T full mesh any site to any other site without traversing the data center routers is required to prevent transcoding and to prevent loose of QoS markings.
2) Voice guy's assertion is that our company's 1Gig point to point links to major sites design for high speed access to the primary data center are NOT suitable paths for voice call setup and certainly not suitable for the RTP to RTP conversational payload.
3) Please review the drawing below and educate me as to where my following understand is not accurate or incomplete.
a. My understanding is voice (regardless hard phone or soft phone) is just an application with specific (granted tight) tolerances...however if the latency is below x, the udp jitter is below y, and congestion is either not present or managed with a properly configured QoS policy then voice quality should be good.
b. Specific best practices and any prevailing standards bodies documentation would be greatly appreciated

Blue line represents Call Setup with OSPF and BGP

Green line represents station to station voice payload RTP

Red line represents what the "voice guy" requires call setup and station to station RTP to be over

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

Drawing to follow....Drawing Must Be Re-Scrubbed To Make Certain it is compliant with information handling and dissemination policies

::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::

Accomplishing this required point voice and (more problematically) data vlans when in voice mode at destinations out one direction from with OSPF where O IA routes were present and would always be preferred ...then also reserving the direction to cause routes that wanted to travel to the destination via BGP because of the AD of 20 installed at redistribution...

A) The former required starving all OSPF area 5 to area 0 inter area route propagation for a specific prefix at every point where area 5 met area zero across 4 vdcs in one data center Nexus pair and also across an OTV link into a secondary data center facility.
B) The later required modifying the OSPF metric from 20 to 19 on a prefix by prefix basis to achieve the demanded massaging.